01
Set it up in NoBonk
Start in the camera view. Test what your phone detects, adjust sensitivity, and choose your sound, vibration or voice cues in a safe space.
NoBonk.Join Android testing BY KRISHAV · HIGH SCHOOL FRESHMAN
A heads-up.
Even outside
the app.
NoBonk is a high school freshman’s effort to help reduce the risks of distracted walking.
Start NoBonk, then switch apps. On Android, its rear-camera AI can keep detecting nearby obstacles and bring alerts onto your screen.
An extra nudge to look up. Never a replacement for watching where you’re going.
REAL ANDROID RECORDING 42 SEC
A student idea with momentum
1st place STEM4All 2026
IEEE Award STEM4All 2026
3rd in category ACSEF 2026
01 / The idea
A school hallway.
A very real question.
Krishav saw students at school bump into each other—or into walls—while looking at their phones. He wondered: could the phone help people notice what was in front of them, even when they were using another app? That is why Android background detection is at the heart of NoBonk.
02
Switch to background
On Android, choose Run in background and allow the requested permissions. Detection can continue while the home screen or another app is visible. Keep the rear camera uncovered and pointed toward the scene.
03
Get a nudge to look up
Alerts can appear above other apps. Use Open NoBonk to return to the controls. Detection can miss hazards, especially in poor light or when the camera points away.
A REAL PROBLEM. A LITTLE WAY TO HELP.
A real problem.
A small idea to help.
Krishav saw students at school bump into each other and walls while looking at their phones. Research on distracted walking shows this is a wider, real-world problem too. NoBonk is his small attempt to help: a nudge to look up when your attention is on your screen.
To cross while texting
In a study of 1,102 pedestrians at 20 selected high-risk Seattle intersections in 2012, texting was associated with an extra 1.87 seconds to cross the average intersection, compared with undistracted walking.
Thompson, Rivara, Ayyagari & Ebel · Injury Prevention, 2013A broader pattern of distraction
A 2020 meta-analysis combined 14 experimental studies. Texting reduced looking left and right before or during street crossing. Phone conversations and texting also increased hits and close calls in the studied settings.
Simmons, Caird, Ta, Sterzer & Hagel · Injury Prevention, 2020A little help with something that matters. These street-crossing studies help explain why distracted walking deserves attention. Thank you to the researchers credited above for helping us understand the problem. NoBonk is Krishav’s contribution: an experimental extra cue to notice what’s nearby. Always look up and stay aware of your surroundings.
THE PERSON BEHIND THE PROJECT
A real question.
His own voice.
From an idea at school to an app on a phone. Hear Krishav explain why he started, then watch him try what he built.
THE APP, IN HIS OWN WORDS
Take a look with Krishav.
A hands-on Android walkthrough: settings, cues and a demonstration of alerts while another app is open.
Recorded September 2026. Background demonstration starts around 1:14. This recording shows a test build; controls can change.
FROM THE PROJECT ARCHIVE
Before the app, a question.
Krishav explains the school-hallway problem, the ideas he considered, and why he wanted the phone itself to help.
Original project presentation, February 2026. It describes early goals and stronger distance/collision claims, not verified guarantees for today’s app. Distance is approximate; alerts can miss hazards.
YouTube loads only after you choose to play. Its privacy policies then apply. Only one of these videos plays here at a time.
FOREGROUND / YOUR SETUP & TEST SPACE
Find your settings.
Then take it with you.
The Android camera view lets you see what your phone detects and tune the cues before switching to background mode. These real Pixel 9a frames show a dog detection and the resulting alert.
These are frames from the actual app, with plain side margins. The distance shown is an estimate. NoBonk can miss or misidentify hazards.
ONE IDEA. TWO DIFFERENT PLATFORMS.
Android today.
iPhone taking shape.
Set up in the app.
Detect beyond it.
Use the foreground camera view to test detections and tweak your settings. Background mode is the main Android experience: user-started camera detection and alerts while another app is open.
- On-device obstacle detection
- Background alerts and an Open NoBonk control
- Camera view, sensitivity and cue settings
Your camera.
Room to do more.
Browse & scan keeps a compact camera view above a website inside NoBonk. Read a page, try supported inline video, or switch to Read / write for a message you choose to paste or write, with the camera still in view.
- People mode and optional Fast Objects detection
- Visible camera, sound and haptic cues
- Browse & scan, plus local reading and writing
Being tested on real iPhones. Public App Store and TestFlight availability are still to come.
iPHONE / A DIFFERENT WAY TO LOOK UP
A view of your path.
A little room for your world.
iPhone works differently from Android. So we’re building an experience around a camera you can see, with useful space beneath it. This is a development preview, not a public release.
NOBONK BROWSE & SCANYour camera stays visibleKeep looking up. No alert is not an all-clear.
A website you chooseRead. Watch. Stay aware.
Compatible web content lives inside NoBonk, below the camera.
Stop scanning ALWAYS WITHIN REACH
Layout illustration · not a camera screenshot
A page below your preview.
Choose Instagram web or YouTube, or enter a secure website address. Compatible pages and inline video sit below the camera. Website features and sign-in depend on the service; Instagram login and messaging are not yet verified in this preview. Full-screen video or switching to another app pauses scanning.
People, or a wider view.
People mode uses Apple Vision. Fast Objects adds eight classes: people, bicycles, cars, motorcycles, buses, trucks, cats and dogs. Both process camera frames on your phone. Neither detects every hazard, and the iPhone preview does not measure distance. Clear status labels distinguish preparation, recent detections and delayed analysis. Old results fade from view while a slower phone keeps processing.
Your words below.
Your camera above.
Choose Read / write. Read mode displays text you explicitly paste in a scrollable view without the keyboard; Write mode lets you compose beneath the camera. Paste adds to your existing text. When you’re ready, stop somewhere safe and choose Pause & share. Apple’s share sheet offers available apps, which may include Messages or WhatsApp. On a configured iPhone, the larger layout also offers a direct Messages composer.
You choose who receives it and confirm sending. Sharing pauses the scan; return and tap Start to resume. Your draft stays if you cancel. NoBonk does not save or sync drafts, and cannot read native Messages, WhatsApp or Instagram inboxes.
Set up while standing still.
Adjust cue sensitivity, Sound and Haptics in Set up. Switch between detection modes after stopping. Use Browse for websites or Read / write for selected text. The camera and Start/Stop control stay separate from the content below; the compact draft layout leaves room for the keyboard.
Return quickly. Start deliberately.
Open NoBonk through Siri or a Shortcut. Supported iPhones can also use an Action button, Control Center or Lock Screen control. These open the app; they do not bypass the reminder or secretly start scanning.
Your camera stays local.
Camera frames are processed in memory, never saved or uploaded by NoBonk. Websites you choose connect to the internet and follow their own privacy policies. The browsing pane uses temporary storage and denies website camera, microphone and motion access.
What to expect today
Keep NoBonk open and the rear camera uncovered. Leaving the app, locking the phone, an interruption or full-screen web media stops scanning. Browsing is not native iPhone split-screen, and Android-style background camera scanning is not included. NoBonk can miss or mislabel objects, especially in poor light. Stay aware of your surroundings.
Can I open Instagram or WhatsApp while scanning?
You can try Instagram’s website inside Browse; sign-in and feature compatibility vary. Opening the native Instagram or WhatsApp app pauses NoBonk. WhatsApp Web is not a supported mobile mode here. Use Read / write to paste selected text or compose beside the camera, then switch to Write and share through an available installed app. It is not a live inbox.
How can I try the iPhone version?
The preview is currently installed directly on development test phones. There is no public iPhone download or TestFlight link yet. Android’s Google Play tester link is for Android only. Follow the project for the public iPhone release.
Does “Fast” mean it will be faster on every iPhone?
Fast is the model’s name. The preview prefers Core ML where available and can fall back to the CPU. Work and temperature affect analysis pace. We are testing newer and older device behavior; we don’t claim that a processor label guarantees better performance or detection.
02 / Inside the decision
Don’t just read it.
Move the signal.
Make a person’s detection box grow. Add an approach signal. Watch the alert change using Android’s actual person-alert thresholds.
DETECTION LABILLUSTRATIVE INPUT
PERSON35% frame height
0%FRAME HEIGHT100%
PERSON / REFERENCE PRESET 2 mLOW
You control the input
Smaller in frameLarger in frame
The tracker sees a closing object.
On-bearing, estimated contact ≤ 1.5 s.
The decision
low alert.
At 28% frame height and above, the person ladder reaches LOW.
Educational simulation of Android’s person alert policy at the 2 m sensitivity preset. No camera access. This is not a distance measurement; the full app also filters detections and applies tracking and alert timing.
Read the actual logic03 / Out of the lab
Real experiments.
Honest numbers.
Early results reported by the project on a Pixel 9a. Promising signals, with plenty still to learn. These are prototype observations, not independent validation.
Detection in good light
Reported detection accuracy under good lighting.
Approach trials correct
Reported approach-detection trial results.
Distance error at 1 m
Distance estimates depend on angle and scene.
The next challenge
Darkness changes the picture.
The project reports 3/10 detection in low light and about 10% battery drain per hour in background mode. Older phones can lag. NoBonk is a student-built prototype that can miss hazards. Keep looking up.
Concept illustration / not an app screenshot04 / On your phone. By design.
A little more awareness.
A lot less oversharing.
Your camera frames stay fleeting.
Processed in memory, then discarded. No saved photos or video. The Android app has no Internet permission. The iPhone preview also keeps camera processing local; websites opened in its browsing pane use the internet.
Your history stays local.
On Android, detection events stay in private storage on your phone and can be cleared. Approximate location is optional and off by default. The iPhone preview does not include this history feature.
IN DEVELOPMENT · OPTIONAL ANDROID FEATURE
Start walking.
Let scanning follow.
We’re building a setup option that can start background scanning after your phone detects sustained walking. You choose whether to enable it—it will be off by default.
PLANNED SETUP OPTION
Walking-triggered background scan
- Choose it in setup. Allow motion access if you want to try it.
- Arm a session in NoBonk. A visible notification will show when it is waiting for walking. The camera stays off while waiting.
- Walk for a while. Sustained walking can start scanning during that session. Stop ends both waiting and scanning.
Still being built and tested. This option is not available in the current Google Play test. For now, open NoBonk and choose Run in background yourself. Android requires a user-started session; this won’t silently start the camera from a stopped app.
A STUDENT’S IDEA. YOUR CHANCE TO HELP.
Built by Krishav.
Better with you.
After seeing fellow students bump into people and walls while looking at their phones, Krishav wondered: could the phone give them a heads-up?
NoBonk grew from that question. On Android, it can scan in the background and alert while another app is open. Help test it on your phone and tell Krishav what works—and what needs work.
Take a closer look at his project board
CLOSED TEST LIVE · INSTALL ON GOOGLE PLAY
Join the Android test
- Join our Google Group. Use the same Google account you use in the Play Store.
- Opt in and install. Use the Google Play testing link below with the same account, install NoBonk, and stay opted in for at least 14 days while trying it and sharing feedback.
Free · Android 10+ · Ages 13+ · Joining the group alone does not start the Google Play testing period. Opt in on Google Play and install the app.
Coming next: an optional walking-triggered setup mode is in development. The current test uses manually started background scanning.
Read the testing instructionsSignup happens on Google Groups. Your account is used to manage test access and updates; the member list is restricted to group owners and managers. NoBonk is a prototype that can miss hazards—keep looking where you’re going.
A FEW MINUTES. USEFUL FEEDBACK.
Your first NoBonk test.
Android 10 or newer, a rear camera, and a safe, well-lit space. The closed test is open. Join the group, then opt in and install through Google Play using the same account.
Use your own Google account
Join the tester group with the account you use in Google Play. A borrowed Android phone is fine: the person testing should sign in with their own account. One person is one tester.
Install through the closed test
Follow the Google Play testing link above, choose to become a tester, then install from Play. Group membership, an APK download, or internal testing alone does not qualify.
Try it in a safe space
Open NoBonk, follow onboarding, and allow the camera when prompted. Point the rear camera at a person or ordinary object with plenty of room around you. Compare the labels and alerts with what you see.
Test background, return & Stop
Try Fast and Sharp modes, sound, vibration, and optional spoken alerts. Choose Run in background and switch to the home screen. Look for the indicator and use Open NoBonk to return. Try Stop in the app and notification: report if the camera indicator, overlays or alerts continue. Check that history can be cleared.
Keep trying it over 14 days
Stay opted in continuously for at least 14 days, install updates, and use the app regularly. We need at least 12 participating testers. Opting out interrupts your consecutive period.
Tell us what happened
Send your phone model, Android version, app version, what you tried, and what you expected. Missed alerts and confusing screens are useful feedback too. Please leave personal details out of screenshots.
Email NoBonk supportKeep looking up. NoBonk is an experimental prototype and can miss hazards. Do not test in traffic, while cycling or driving, or by approaching anything dangerous.
How do I update NoBonk?
Keep the same Google account enrolled in the closed test. Open NoBonk in Google Play and choose Update when a newer version is available to you. If NoBonk offers an in-app update, you can choose Update or Later; apply a downloaded update when scanning is stopped. Your first installation of a version with update prompts still needs to come through Play normally.
Already installed a test APK, or joined internal testing?
A directly downloaded APK does not start Google’s closed-test period. An earlier debug-signed APK may need to be uninstalled before the Play version will install; uninstalling removes its local history. If you joined internal testing, leave that test before opting into the closed test.
Google Groups says “Content unavailable”?
Open the group in Chrome and check the Google account shown at the top right. Use the same account as Google Play, then try the group link again. If it still fails, email support with the account you want to use; do not send your password.
THE PEOPLE & TOOLS BEHIND NOBONK
Krishav’s idea.
Plenty of help along the way.
Created by Krishav
Krishav identified the school-hallway problem, shaped the app’s purpose and behavior, and tested it on real phones.
Haarith (Dad) — Thank you for believing in this idea and sitting with me through the frustrating parts, especially when nothing seemed to work. Having you there made it easier to keep going. And thank you, Mom and Dad, for faithfully paying the increasingly ridiculous AI bills without asking too many questions :)
Kaaval & the local model team
Over the months of development, Kaaval—the family’s local AI coding setup—helped with coding and build-and-test iterations. It runs across multiple NVIDIA DGX Spark systems, using a mix of Qwen, DeepSeek and GLM 5.x models.
AI-assisted, openly credited
Claude, OpenAI Codex and ChatGPT Astra also helped with implementation, review, the website and release work. Google Gemini and Warp supported the broader workflow. These tools contributed substantially; Krishav guided the project and evaluated the results.
This describes the tools used to build NoBonk. The Android app’s camera processing runs on your phone; it does not send camera frames to Kaaval or these AI services.
Project acknowledgments on GitHubTHE ANDROID CLOSED TEST IS OPEN
Less bonk.
More world.
A school-hallway question, built into a background companion.
Help Krishav test it on Android.
JOIN THE CLOSED TEST ONGoogle PlayAndroid 10+ · Free closed test · iPhone development preview · public release coming soon Follow the project on GitHub